UDA-LIDI/openai-whisper-large-v3-fullFT-es_ecu911_V2martin_seg30s_samples

Whisper openai-whisper-large-v3

This model is a fine-tuned version of openai/whisper-large-v3 on the llamadas ecu911 dataset. It achieves the following results on the evaluation set:

• Loss: 1.4269
• Wer: 63.4503

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.027	0.9860	53	0.9172	56.4327
0.528	1.9907	107	0.9384	53.5088
0.2863	2.9953	161	1.0114	60.5263
0.1576	4.0	215	1.1557	65.1072
0.0986	4.9860	268	1.1991	64.1326
0.0639	5.9907	322	1.1858	54.3860
0.048	6.9953	376	1.2570	57.0175
0.0368	8.0	430	1.2571	56.2378
0.0341	8.9860	483	1.2981	68.0312
0.0257	9.8605	530	1.4269	63.4503

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1